HAL Id: ineris-00961872
https://hal-ineris.archives-ouvertes.fr/ineris-00961872
Submitted on 20 Mar 2014
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Modelling air quality in street canyons : a review
Sotiris Vardoulakis, Bernard E.A. Fisher, Koulis Pericleous, Norbert
Gonzalez-Flesca
To cite this version:
Sotiris Vardoulakis, Bernard E.A. Fisher, Koulis Pericleous, Norbert Gonzalez-Flesca. Modelling air
quality in street canyons : a review. Atmospheric environment, Elsevier, 2003, 37 (2), pp.155-182.
�10.1016/S1352-2310(02)00857-9�. �ineris-00961872�
MODELLING AIR QUALITY IN STREET CANYONS: A REVIEW
Sotiris Vardoulakis
a'
c, Bernard E.A. Fisher , Koulis Pericleous
a'*, Norbert Gonzalez-Flesca
c"School of Computing and Mathematical Sciences, University of Greenwich, Maritime Greenwich University Campus, 30 Park Row, London SEW 9LS, UK
'' National Centre for Risk Analysis & Options Appraisal, Environment Agency, Kings Meadow House, Kings Meadow Road, Reading RG1 8DG, UK
' Institut National de l'Environnement Industriel et des Risques (INERIS), Parc Technologique ALA TA, BP 2, 60550
Verneuil-en-Halatte, France
Abstract
High pollution levels have been often observed in urban street canyons due to the increased traffic emissions and reduced natural ventilation. Microscale dispersion models with different levels of sophistication may be used to assess urban air quality and support decision making for pollution control strategies and traffic planning. Mathematical models calculate pollutant concentrations by solving either analytically a simplified set of parametric equations or numerically a set of differential equations that describe in detail wind flow and pollutant dispersion. Street canyon models, which might also include simplified photochemistry and particle deposition-resuspension algorithms, are often nested within larger-scale urban dispersion codes. Reduced-scale physical models in wind tunnels may also be used for investigating atmospheric processes within urban canyons and validating mathematical models.
A range of monitoring techniques is used to measure pollutant concentrations in urban streets. Point measurement methods (continuous monitoring, passive and active pre-concentration sampling, grab sampling) are available for gaseous pollutants. A number of sampling techniques (mainly based on filtration and impaction) can be used to obtain mass concentration, size distribution and chemical composition of particles. A combination of different sampling monitoring techniques is often adopted in experimental studies. Relatively simple mathematical models have usually been used in association with field measurements to obtain and interpret time series of pollutant concentrations at a limited number of receptor locations in street canyons. On the other hand, sophisticated numerical codes have often been applied in combination with wind tunnel and/or field data to simulate small-scale dispersion within the urban canopy.
Keywords: Urban environment; Air pollution; Street canyon; Dispersion models; Computational Fluid Dynamics (CFD);
Wind flow regimes; Traffic emissions
Corresponding author: Professor K. Pericleous
E-mail: [email protected], Tel: +44 (0) 208 3318732, Fax: +44 (0) 208 33 18695
1. Introduction
/. /. Urban air quality and traffic
Urban air pollution was originally considered as a local problem mainly associated with space heating and industrial emissions, which are now controllable to a great extent. Despite significant improvements in fuel and engine technology, present day urban environments are mostly dominated by traffic emissions (Fenger, 1999; Colvile et al., 2001). It is now generally recognised that many of the substances directly emitted by vehicles in the ambient air or indirectly produced through photochemical reactions represent a serious hazard for human health (Hoek et al., 2000; Nyberg et al., 2000; Dab et al., 2001).
The main traffic-related pollutants are CO, NOX, hydrocarbons, and particles. CO is an imperfect fuel combustion
product. Combustion also produces a mixture of N O2 and NO, of which more than 9 0 % is in the form of NO. A wide
range of unburned and chemically transformed hydrocarbons (e.g. benzene, toluene, ethane, ethylene, pentane, etc.) is emitted by motor vehicles through a number of different processes (e.g. evaporation, fuel tank displacement, oil seep, etc.). Finally, particles of condensed carbonaceous material are emitted mainly by diesel and purely maintained petrol vehicles.
Atmospheric pollutants are responsible for both acute and chronic effects on human health (WHO, 2000). CO is an asphyxiating pollutant that reduces the ability of blood to carry oxygen to the different organs (Burnett et al., 1998). Therefore, short-term exposure to high CO concentrations might cause an acute health impact. On the other hand, pollutants like benzene have an accumulative effect on human health. Long-term exposure to high benzene levels increases the risk for an individual to suffer from cancer (Cicolella, 1997). Furthermore, there are gases like NO2 that are responsible for both short- and long-term health effects. Depending on the effects related to each substance, atmospheric pollutants are regulated with respect to different exposure times. For example, in the European air quality guidelines, standards are set for benzene as one year averages, for CO as eight hour averages, and for N O2 as both one hour and one year averages (European Commission. 1999; 2000).
Particulate matter with aerodynamic diameter below 10 u.m (PM|0) and especially the finer fraction with aerodynamic diameter below 2.5 um (PM2.5) was found to associate with increased daily mortality and asthma (Dockery and Pope, 1994; Anderson at al.. 1992; Harrison and Yin, 2000). Nevertheless, current European legislation addresses only total PMio as 24-hour and one year averages, while U.S. legislation regulates both PMi 0 and P M2 5 as three year averages (EPA, 1996).
Although roadside concentrations differ significantly from background levels, all outdoor environments are subject to the same regulatory standards for ambient air quality. This is due to the assumption that pollutant concentrations observed at a single or a few permanent monitoring stations within a city are representative of the exposure of the entire urban population.
In urban environments and especially in those areas where population and traffic density are relatively high, human exposure to hazardous substances is expected to be significantly increased. This is often the case near busy traffic axis in city centres, where urban topography and microclimate may contribute to the creation of poor air dispersion conditions giving rise to contamination hotspots. High pollution levels have been observed in street canyons, which is a term frequently used for urban streets flanked by buildings on both sides. Within these streets, pedestrians, cyclists, drivers and residents are likely to be exposed to pollutant concentrations exceeding current air quality standards.
1.2. Air quality monitoring
Nowadays, automated monitoring networks operate in many European cities providing detailed air quality information on a regular basis. There are several techniques available for monitoring gaseous pollutants (e.g. continuous monitoring using standard gas analysers, diffusive and pumped sampling using tubes filled with an appropriate adsorbent, grab sampling using canisters) and particulate matter (e.g. filtration and impaction). Each one of them can be associated with a number of advantages and disadvantages that make it suitable or not for a specific application.
The response time, which is the time over which the sample is taken, is one of the major factors that determine the suitability of a method. Standard gas analysers are sufficiently sensitive and fast to give real time (i.e. typical response time: 1-2 min) measurements of CO, NOX and O3 concentrations. The results can be then averaged over short time periods and compared to the regulatory standards.
Diffusive samplers have a relatively long response time (e.g. typically from one/two days to one/two weeks), which makes them less suitable for observing atmospheric pollutants responsible for short-term health effects. On the other hand, long response times might be preferable when sampling substances like benzene, whose impact on human health is due to cumulative exposure. In these cases, peak concentrations are of minor concern and therefore diffusive samplers appear to be the ideal choice (Cocheo et al., 2000; Skov et al., 2001). Furthermore, diffusive samplers are portable devices and do not need electrical power supply, which makes them very suitable for spatial distribution measurements (e.g. vertical profiles within canyons), air quality mapping, personal exposure studies, and detection of long-term pollution trends.
The total number of air quality monitoring stations or sampling locations within a city is limited by practical constraints. Since pollutant concentrations might vary with a factor of 5 from a street canyon to an urban background area (Palmgren and Kemp, 1999), the selection of monitoring/sampling locations becomes fundamental. In addition, one should be cautious when comparing absolute values from different cities. The data are often based on one or few monitoring stations placed at critical sites and thus represent microenvironments rather than large urban areas (Fenger, 1999). In general, monitoring stations and/or samplers should be located near places of expected air pollution hotspots, but also must be reasonable with respect to population exposure over the averaging times associated with the regulatory values.
Permanent air quality stations within a city may be classified into two broad categories: (a) the roadside and (b) the urban background stations. Roadside stations are usually located on the pavement of busy streets, avenues or intersections, within few meters distance from the roadway and with their sampling head at 1.5 - 3 m height above ground. On the other hand, background stations are placed in parks or other urban locations away from road traffic.
1.3. Air quality model Iing
The impact of air pollution on urban climates has become an important research issue (Georgii, 1969; Oke, 1988; Bitan, 1992), leading to numerous modelling studies related to the influence of buildings and other urban structures on pollutant accumulation/dissipation patterns. The main features of pollutant dispersion within urban canyons are well understood through the pioneering work of Johnson et al. (1973), Dabberdt et al. (1973), Hotchkiss and Harlow (1973), Nicholson (1975). Yamartino and Wiegand (1986) and others.
Dispersion models are now widely used for assessing roadside air quality by providing predictions of present and future air pollution levels as well as temporal and spatial variations (Sharma and Khare, 2001). When used in a knowledgeable way. they can be very useful in giving insights into the physical and chemical processes that govern the dispersion and transformation of atmospheric pollutants. This article provides an overview of modelling techniques and available software for assessing air quality in street canyons. In this framework, relevant experimental and theoretical studies are also briefly discussed.
2. Street canyon characteristics
2.1. Canyon geometry
The term street canyon ideally refers to a relatively narrow street with buildings lined up continuously along both sides (Nicholson. 1975). However, the same term has been used to refer to larger streets, also called avenue canyons. In the real world, a broader definition of the term has been applied, including urban streets that are not necessarily flanked by buildings continuously on both sides, allowing thus for some openings on the walls of the canyon.
The dimensions of a street canyon are usually expressed by its aspect ratio, which is the height (//) of the canyon divided by the width (II). A canyon might be called regular, if it has an aspect ratio of approximately equal to 1 and no major openings on the walls. An avenue canyon may have an aspect ratio below 0.5, while a value of 2 may be representative of a deep canyon. Finally, the length (L) of the canyon usually expresses the road distance between two major intersections, subdividing street canyons into short (L/H~3), medium (L/H~5), and long canyons ( £ / / / « 7). Urban streets might be also classified in symmetric (or even) canyons, if the buildings flanking the street have approximately the same height, or asymmetric, if there are significant differences in building height.
2.2. Wind flow
The climate of street canyons is primarily controlled by the micro-meteorological effects of urban geometry rather than the mesoscale forces controlling the climate of the boundary layer (Hunter et al., 1992). A clear distinction should be made between the synoptic above roof-top wind conditions and the local wind flow within the cavity of the canyon (Fig. 1). Depending on the synoptic wind (or free-stream velocity), three main dispersion conditions can be identified: (i) low wind conditions, for synoptic winds lower than 1.5 m/s, (ii) perpendicular or near-perpendicular flow for synoptic winds over 1.5 m/s blowing at an angle of more than 30° to the canyon axes, and (iii) parallel or near-parallel flow for winds over 1.5 m/s blowing from all other directions. In the case of perpendicular flow, the upwind side of the canyon is usually called leeward, and the downwind windward.
The emphasis has often been on the two-dimensional nature of the flow, studying vertical cross-sections at mid-canyon level. When the above roof flow is perpendicular to the canyon and the wind speed is greater than 1.5 to 2 m/s, flow may be described in terms of three regimes, depending on the dimensions of the street (Oke, 1988) (Fig. 2): (a) isolated
roughness flow, (b) wake interference flow, and (c) skimming flow. For wide canyons (H/W<03), the buildings are well
spaced and act essentially as isolated roughness elements, since the air travels a sufficient distance downwind of the first building before encountering the next obstacle. As buildings become more closely spaced {H/W~ 0.5), the disturbed air flow has insufficient distance to readjust before encountering the downwind building, resulting in wake interference flow. In the case of regular canyons (H/W~ 1), the bulk of the synoptic flow skims over the canyon producing the
From a three-dimensional point of view, a reflection of the wind off the windward wall of the canyon should be ideally observed in the case of skimming flow (Nakamura and Oke, 1988; Johnson and Hunter, 1999). For oblique roof-level winds, this reflection may induce a spiral wind flow through the canyon. Other complex channelling effects might be produced for winds parallel to the street axis. Additional low pressure areas and wind circulation is created near intersections, resulting in horizontal corner vortices. In relatively short canyons, corner vortices might be strong enough to inhibit a stable vortex perpendicular to the street in the mid-section. This ventilation effect fades with increasing street length (Theurer. 1999).
The strength of the wind vortices inside the canyon mainly depends on wind speed at roof-top level. However, the local wind flow is also affected by the mechanical turbulence induced by moving vehicles (Eskridge and Rao, 1986) or created near urban roughness elements within the street (e.g. trees, kiosks, balconies, slanted building roofs, etc.) (Hoydysh and Dabberdt, 1994; Theurer. 1999). Furthermore, the shape and strength of the wind vortices might be also affected by the atmospheric stability and other thermal effects induced by the differential heating of the walls and/or the bottom of the canyon (Sini et al., I996; Kim and Baik, 2001).
In relatively deep canyons (H/W>\3), the main wind vortex is usually displaced towards the upper part of the cavity, with almost stagnant air below (DePaul and Shieh, 1986). As the aspect ratio increases (H/W~2), a weak counter-rotating secondary voitex may be observed at street level (Pavageau et al., 1996). For even higher aspect ratios
(H/W~3), a third weak vortex might be also formed (Jeong and Andrews, 2002). In most cases, small week vortices
occupy the bottom side corners of the canyon.
Depending on the wind direction, asymmetric canyons may be sub-divided into two categories: (i) step-up canyons, when the down-wind building is higher than the up-wind building, and (ii) step-down canyons, when the down-wind building is lower than the up-wind building. In these cases, mid-section wind vortices might be displaced or reversed within the cavity.
2.3. Pollutant dispersion
The dispersion of gaseous pollutants in a street canyon depends generally on the rate at which the street exchanges air vertically with the above-roof level atmosphere and laterally with connecting streets (Riain et al., 1998). Skimming flow, a feature of regular canyons, provides minimal ventilation of the canyon and is relatively ineffective in removing pollutants (Hunter et al., 1992). Field measurements (DePaul and Sheih, 1985, Qin and Kot, 1993) show increased concentrations of traffic-related pollutants on the leeward side of the canyon, and decreasing concentrations along with height above the ground on both sides of the street. The increased leeward concentrations are due to the accumulation of pollutants locally advected by the large wind vortex that covers most of the canyon. Minor pollution hotspots might be also created in small cavities where additional recirculation phenomena can take place.
Street-level cross-road gradients observed in wind tunnel experiments (Hoydysh and Dabberdt, 1988) for perpendicular wind conditions indicate that concentrations are generally a factor of two greater for the leeward than for the windward side, except for step-down canyons where windward concentrations are slightly greater than leeward. Concentrations are generally lower in the step-up canyons relative to the even and step-down notches.
Flow visualisation experiments have shown that the rotation of the main wind vortex in the canyon is periodically interrupted (Hoydysh and Dabberdt, 1988; Pavageau et al., 1996). As a result, the pollutants trapped within this vortex are periodically flushed out of the canyon, a phenomenon known as canyon breathing (Scaperdas, 2000). In relatively long canyons without connecting streets, maximum street-level concentrations are more likely to occur when the synoptic wind is parallel to the street axis. In that case, the accumulation of emissions along the line source outweighs the ventilation induced by the parallel winds (Soulhac et al., 1999; Dabberdt and Hoydysh, 1991).
Low synoptic winds create a well-known meteorological situation that favours air pollution built-up in urban areas (Qin and Kot, 1993; Vignati et al., 1996; Jones et al. 2000). When the synoptic wind speed is below 1.5 m/s, the wind vortex within the canyon disappears and the air stagnates in the street. In that case, the mechanical turbulence induced by moving vehicle as well as the atmospheric stability conditions might play a significant role in the dispersion of traffic-generated pollutants.
Fine and especially ultra-fine particles are expected to disperse in the air like gases. The larger-sized particles, however, are greatly affected by gravity and thus have a shorter residence time in the air (Chan and Kwok, 2000). For this reason,
the coarse fraction of the total suspended particles (TSP) exhibits larger vertical concentration gradients than those usually observed for gases or fine particles.
2.4. Pollutant transformation
Due to the very short distances between sources and receptors, only very fast chemical reactions have a significant influence on the measured concentrations within street canyons (Berkowicz et al., 1997). For this reason, most traffic-related pollutants (e.g. CO and hydrocarbons) can be considered as practically inert species within these distances. This is not the case either for NO2, which dissociates extremely fast in the presence of light, or for NO, which also reacts very fast with O;,. The time scales of these chemical reactions are of the order of tens of seconds, thus comparable with residence times of the pollutants in a street canyon.
It is expected that the relationship between relatively stable chemical species emitted by vehicles would not vary significantly within urban streets. This is very helpful for epidemiological studies, because a single or only few indicators can be identified for assessing population exposure to traffic-related air pollution (Kingham et al., 2000).
2.5. Population exposure
From a population exposure point of view, air quality in street canyons is of a major importance, since the highest pollution levels and the larger targets of impact are often concentrated in this kind of streets. The so-called canyon effect (i.e. the reduced natural ventilation in urban streets) results in greater health impacts (e.g. indicated by an increased number of respiratory hospital admissions) and damage costs for the exposed population (Spadaro and Rabl, 2001). Personal exposure can be calculated as the product of the pollutant concentration and time spent in a specific
microenvironment, which is defined as a confined space (e.g. bedroom, office, car, parking, pavement, etc.) where
pollutant concentrations are assumed to be uniform (Colls and Micallef, 1997). The total personal exposure will be then the sum of all such products. However, the assumption of spatial uniformity of air pollution might be erroneous for certain microenvironments like street canyons, where strong spatial concentration gradients are often observed. In these cases, exposure calculations should be refined by subdividing microenvironments into sub-microenvironments, taking into account pollution hot spots and refined human breathing zones (e.g. for residents, pedestrians, cyclists, drivers, etc.). Relatively few examples of this approach can be found in literature. In a study attempting to quantify residential exposure to exhaust gases in Oslo (Larssen et al., 1993), a correction coefficient was introduced to account for changes in ambient concentrations with height over street level. Other authors (Croxford and Penn, 1998) have suggested that a side of the street factor should be introduced if the prevailing wind direction is perpendicular or near-perpendicular to the street axis.
3. Street canyon modelling
There is a plethora of dispersion models specially developed for or simply used in street canyon applications. They can be useful in air quality and traffic management, urban planning, interpretation of monitoring data, pollution forecasting, human exposure studies, etc. Although there are no clear-cut distinctions between different categories, models might be classified into groups according to their physical (e.g. reduced-scale) or mathematical principles (e.g. box, Gaussian, CFD) and their level of sophistication (e.g. screening, semi-empirical, numerical). Some of these (often overlapping) categories and corresponding models are presented in Table 1.
3.1. Parametric models 3.1.1. Gaussian plume models
These are sets of equations describing the three-dimensional concentration field generated usually by a point source. They assume that the concentrations from a continuously emitting source are proportional to the emission rate, inversely proportional to the wind speed, and that the time averaged pollutant concentrations horizontally and vertically are well described by Gaussian (i.e. bell-shaped) distributions (Boubel et al., I994). In its simplest form, the Gaussian plume model assumes that there are no chemical or removal processes taking place and that pollutant material reaching the ground or the top of the mixing layer as the plume grows is reflected back towards the plume centreline.
Gaussian plumcmodels rely on the appropriate selection of the plume spread sigma functions (in both the horizontal and vertical sense), which are generally expressed in terms of Pasquill atmospheric stability classes or Monin-Obukhov similarity theory parameters (Zannetti, 1990). Models using the latter approach dispose of height dependent sigma functions and are known as second generation Gaussian plume models (Carruthers et al., 1994).
Apart from industrial applications (i.e. point sources), specially designed Gaussian plume models can be used to calculate pollutant concentrations over urban agglomerations (i.e. area sources) and in the vicinity of highways (i.e. line sources). Gaussian models are not directly applicable to small-scale dispersion within the urban canopy, since they treat buildings and other obstacles only via a surface roughness parameterisation (Scaperdas, 2000). Nevertheless, in some cases, they include specialised modules for street canyons. This is the case of ADMS-Urban (Owen et al., 1999), a second generation urban-scale dispersion model that includes a street canyon module nested within the core Gaussian code.
3.1.2. CALINE4
CALINE4, the latest version of the CALINE series of pollutant dispersion models, is one of the most validated models available for assessing the impact of vehicle traffic on roadside air quality (Benson, 1984). It has been widely used in scientific and engineering applications mainly concerning highway development and management (Jones et al, 2000). Although CALINE4 is able to handle canyon or intersection situations, it has been used in relatively few urban air quality studies.
The model uses Gaussian plume theory to simulate the dispersion of pollutants emitted from a line source. This is divided in a series of elements, which are modelled as equivalent finite line sources located normal to the wind direction. The region directly over the road, called the mixing zone, is treated as a zone of uniform emission and turbulence. Within the mixing zone, vehicle induced turbulence (both mechanical and thermal) is taken into account (Benson, 1992).
The street canyon algorithm devised by Turner (1970) computes the effect of single or multiple horizontal reflections of the plume on the walls of the canyon. In this case, the road axis and the wind velocity are assumed to be parallel to the horizontal topographic boundary (i.e. the walls of the canyon), resulting in equal concentrations on both sides of the street.
3.1.3. TW and CAR
TNO-Traffic is a Gaussian plume-type model that describes the dispersion of traffic exhausts (Eerens et al., 1993). It is based on an extensive programme of wind tunnel experiments which covered a great number of different street configurations, including urban canyons and intersections (van den Hout et al., 1994). In this model, the traffic is represented by line sources divided into series of small point sources.
CAR (or CAR-International) is a simplified version of the same model, in which only the most representative street configurations were categorised (den Boeft et al., 1996). For each street type (e.g. highway, urban road, etc.) a source-receptor relationship is specified as a function of the distance between the source-receptor and the street axis. CAR uses annual average wind speeds and assumes that there is no prevailing wind direction. Thus, the user obtains the same yearly averages and percentiles on either side of the street. In all cases, the effect of trees and moving vehicles on street-level wind velocity is taken into account. This model should not be confused with CAR-FMI (Harkônen et al., 1995), which is a Gaussian line source model for calculating pollution from road networks.
3.1.4. STREET-SRI
Johnson et al. (1973) used a single box model, together with some simplified assumptions concerning initial dispersion and car induced turbulence, to derive a street canyon sub-model usually called STREET or SRI (e.g. Stanford Research Institute), which formed part of a multipurpose urban diffusion model for inert pollutants (APRAC). It is based on the assumption that concentrations of the pollutant occurring on the roadside consist of two components, the urban background concentration and the concentration component due to vehicle emissions generated within the specific street. Then, it calculates pollutant concentrations on both sides of the street, taking into account the height and distance of the simulated receptor from the kerb.
On the leeward side of the canyon, concentrations are assumed to be inversely proportional to the distance between the line source and the receptor point. On the windward side, the original expression given by Johnson et al. (1973) was revised by Dabberdt et al. (1973) to account for vertical decrease of concentrations due to entrainment of fresh air
through the top of the canyon. For parallel or near-parallel synoptic winds, the average of the leeward and windward values might give the pollutant concentration on both sides of the street, although the model is not specifically designed for this situation. STREET-SRI was parameterised using data from a regular street canyon and for this reason it might need re-calibration before being applied to other canyon geometries.
3.1.5. CPBM
The Canyon Plume Box Model (Yamartino and Wiegand, 1986) combines a Gaussian plume model for the direct impact of pollutants emitted in the street, with a box model that accounts for the additional impact of pollutants trapped within the wind vortex formed inside the canyon. The wind flow in the canyon is reproduced using the methodology proposed by Hotchkissand Harlow (1973) for the two transverse components of the wind velocity and a logarithmic expression for the longitudinal component. An empirical model that takes into account wind generated turbulence as well as thermal effects induced by soiar radiation and moving vehicles is used to calculate the turbulent sigma parameters representing the standard deviation of flow velocities about the mean flow.
The plume generated inside the canyon is divided into three segments, which are assumed to follow straight line trajectories and disperse according to Gaussian plume formulas. The impact resulting from the recirculation component is calculated from the consideration of the mass budget inside the canyon. On the leeward side of the street, the total impact is calculated by adding the direct plume to the recirculated fraction. On the windward side, where the only contribution arises from the recirculation component, the dilution of the concentrations due to the entrainment of fresh air is also taken into account. For winds parallel to the street axis or for very low wind speeds, a simpler plume model is used.
3.1.6. USP.\ 1 and A EOL1US
AEOLIUS (Buckland, 1998) is based on concepts and techniques previously used for the development of the Operational Street Pollution Model (Hertel and Berkowicz, 1989a), which was evolved from the CPBM. AEOLIUS and OSPM are semi-empirical models that calculate concentrations of exhaust gases on both sides of a canyon assuming three different contributions: (a) the contribution from the direct flow of pollutants from the source to the receptor, (b) the recirculation component due to the flow of pollutants around the vortex generated within the recirculation zone of the canyon, and (c) the urban background contribution. A Gaussian plume algorithm is used for the calculation of the direct contribution and a simple box model for deriving the recirculation component.
The vortex is formed inside the canyon, if the synoptic wind is not parallel to the street axis. The length of the vortex (along the wind direction) is 2 times the upwind building height. For synoptic winds below 2 m/s, the length of the vortex decreases with the wind speed (Berkowicz, 2000a). The width of the recirculation zone cannot exceed the width of the canyon in any case. The relation between street- and roof-level winds in the canyon is given by a logarithmic relationship that takes into account the surface roughness length, the height of initial dispersion of car exhausts and the synoptic wind direction.
On the leeward side of the street, concentrations are calculated as the sum of the direct and recirculation contributions, while on the windward side only the direct contribution of emissions generated outside the recirculation zone are taken into account. If the recirculation zone extends throughout the whole canyon, then the windward concentrations are calculated from only the recirculation component. For near-parallel flow, emissions from outside the recirculation zone may contribute to the leeward concentrations. When the wind speed is near zero or parallel to the street axis, the concentrations on both sides of the canyon become equal. Finally, the mechanical turbulence in the street due to the wind and vehicle traffic is empirically derived.
AEOLIUS (the Full version) is based on the same formulation as OSPM. Nevertheless, some discrepancies between predictions from the two models cannot be excluded, due to differences in coding, parameterisation and data pre-processing techniques. There are also two screening versions of AEOLIUS, namely AEOLIUS Screen and AEOLIUSQ Emission, made available by the U.K Met Office.
3.1.7. Receptor models
The models described in the previous sections may be also defined as source-oriented models. Such models rely on the use of best available emission estimates and meteorological data to predict pollutant concentrations at various roadside locations. An alternative approach is the receptor-oriented modelling, which is based on the detailed analysis of the pollutant collected at one or more monitoring sites. This analysis, also called source apportionment or chemical mass
balance (Gordon et al., 1984), attempts to determine which sources contributed to the concentration measured at the
receptor point. If the pollutant of interest is chemically inert (e.g. CO), there is no way to distinguish between different sources. But if the pollutant consists of a variety of chemical species (e.g. participate matter), then from its chemical composition one can make inferences about the sources.
Receptor-oriented models, such as the Constrained Physical Receptor Model (CPRM) (Wâhlin et al., 2001), are mostly used to test the predictions made by source-oriented models as well as the accuracy of the emission estimates that are used in them (Karim and Ohno, 2000).
3.2. CFD models
Computational Fluid Dynamics (CFD) modelling is a general term used to describe the analysis of systems involving fluid flow, heat transfer and associated phenomena (e.g. chemical reactions) by means of sophisticated computer-based numerical methods. It is a powerful modelling technique spanning a wide range of industrial and more recently environmental and biomédical applications (Gosman, 1999).
What distinguishes CFD from other Eulerian models is their capability to deal with very complex shaped walls and other boundary conditions (e.g. in aircraft and automobile design) using flexible fine-scale grids. Furthermore, they include a more sophisticated treatment of turbulence, which makes them suitable for small-scale pollutant dispersion applications. CFD codes are structured around numerical algorithms that can tackle fluid flow problems. In order to provide easy user access, most commercial CFD packages include sophisticated input and output interfaces. Hence, they contain three main elements: (I) The pre-processor, which serves to input problem parameters, generate the grid of the computational domain, select the physical and chemical phenomena that need to be treated, define the fluid properties, and finally specify the appropriate boundary conditions. (II) The solver, which first approximates numerically the unknown flow variables, then discretises the governing flow equations using these approximations, and finally solves the resulting system of algebraic equations. (Ill) The post-processor, which displays the grid and geometry of the domain, plots vectors (e.g. wind velocity) and contours (e.g. pollutant concentration) and may even provide animation facilities for dynamic result display.
3.2.1. Physical principles
CFD modelling is based on the numerical solution of the governing fluid flow and dispersion equations, which are derived from basic conservation and transport principles: (a) the mass conservation (continuity) equation, (b) the three momentum conservation (Navier-Stokes) equations in x, y, z, and (c) the transport equation for pollutant concentration. The equations of state (obtained through the thermodynamic equilibrium assumption) and the Newtonian model of viscous stresses are also enlisted to close the system numerically. The initial and boundary conditions are specified by the user.
Furthermore, atmospheric turbulent processes need to be modelled. Existing turbulence models can be classified in two broad categories: (I) The classical models based on Reynolds Averaged Navier-Stokes (RANS) flow equations (e.g. the k - s model, which is by far the most used and validated); (II) the Large Eddy Simulation (LES) models, which are computationally very demanding and therefore mainly used in research applications (Versteeg and Malalasekera, 1995).
3.2.2. Numerical principles
There are three different streams of numerical solution techniques: finite difference, fine element, and spectral methods. The main differences between them are associated with the way in which the flow variables are approximated and with the discretisation processes. The finite volume method, which was originally developed as a special finite difference formulation, is now the most well established and thoroughly validated method (it is central to most popular CFD codes: PHOENICS. FLUENT, STAR-CD).
According to this method, the flow domain is divided into individual/fa/te control volumes (or computational cells). The differential flow equations are then integrated over each cell in order to transform them into a set of approximated algebraic difference equations between all nodal points of the grid. An advantage of the finite volume method is that mass and momentum conservation is imposed at cell level, which ensures that the discretised form of the flow equations integrated over the entire domain is also conservative.
An iterative approach is required for solving the system of algebraic difference equations resulting from the discretisation method. The most popular solution procedures are the TDMA, a line-by-line solver of the algebraic equations, and the SIMPLE algorithm. The Semi-Implicit Method for Pressure Linked Equations (SIMPLE), originally proposed by Patankar and Spalding (1972), is a predictor-corrector method. That means that velocities are predictedby solving the momentum conservation equations using the most recent estimate of the pressure field, and then the pressure field is
corrected'by using the imbalances in the mass conservation equations. The other conservation equations are then solved,
and the procedure is iterated until reaching convergence (i.e. when the imbalance in all conservation equations reaches a sufficient low value).
3.2.3. Applications and codes
When pollutant dispersion is examined within a street canyon, the computational domain should be sufficiently extended to stabilise the air inflow and outflow through the geometrical boundaries of the area. The relief of the buildings (e.g. due to the presence of balconies) or the street (e.g. due to the presence of vegetation, parked cars, etc.) can be taken into account by introducing a roughness coefficient for each surface of the domain. The space discretisation is usually not uniform, since a higher resolution is required near the canyon walls and the roadway. Finally, fields of pollutant concentrations, wind velocity and other physical quantities (e.g. turbulent kinetic energy and eddy diffusivity) may be reproduced.
The commercially available general-purpose CFD codes PHOENICS, FLUENT, STAR-CD, CFX-TASCflow and Fluidyn-PANACHE have been used in a number of street canyon applications. Other numerical models like MERCURE (Carissimo et al.. 1995). CHENSI (Levi Alvares and Sini, 1992) and MISKAM (Eichhorn, 1995) were specially designed to simulate pollutant dispersion at local scale. MISKAM was used to create a database of numerical three-dimensional simulations that was integrated in a screening model called STREET (Petit et al., 2000). Furthermore, the street canyon module PROKAS-B. which forms part of the Gaussian urban scale model PROKAS-V, was also based on dimensionless concentrations calculated using a version of MISKAM. Finally, the microscale models MIMO and MITRAS were also specially designed for street canyon applications and nested within the mesoscale MEMO and METRAS, respectively (Ehrhardetal.. 2000).
3.3. Reduced-scale models
The reduced-scale (or physical) models are based on the principle that by reducing the geometrical scale of a given flow domain and adjusting the reference parameters (e.g. flow velocity), the original full-scale conditions can be reproduced. Reduced-scale modelling can be carried out in a wind tunnel or a water tank facility. Although wind tunnels have been more widely used for simulating pollutant dispersion than water tanks, the same principles and considerations apply to both the methods.
Three monitoring techniques may be involved in wind tunnel experiments: (a) flow visualisation, which helps to explore the range of possible flow and dispersion patterns obtained for different building arrangements, (b) tracer dispersion, which is used to quantify concentrations at receptor locations within the canyon, and (c) Laser Doppler Anemometry (LDA), which is used to study in more detail the patterns observed during flow visualisation experiments.
Although wind tunnel modelling involves a considerable simplification and idealisation of the real full-scale situation in terms of street geometry, meteorology, traffic flow and emissions, it can efficiently approximate real atmospheric conditions in urban streets. Furthermore, it allows isolating and studying separately each one of the phenomena involved in microscale pollutant dispersion. Wind tunnel measurements have been widely used for model development and validation (Baker and Hargreaves, 2001). However, differences between wind tunnel and full-scale experimental data should be carefully considered when validating numerical models (Schatzmann et al., 1999).
3.4. Model requirements
Dispersion model predictions are in most cases a function of meteorology, street geometry, receptor location, traffic volumes and emission factors. The acquisition and pre-processing of these data is an important part of any modelling study, since the performance of a model greatly depends on the quality of the inputs.
Detailed traffic'information, including traffic volumes, fleet composition (e.g. ratio of light/heavy duty vehicles) and average vehicle speeds, is normally required for running street canyon models. Part of this information (e.g. traffic volumes and average vehicle speeds) might be obtained from automatic detectors permanently or temporarily operating in the street of interest. The vehicle fleet composition, however, is rarely available for a specific location and time period and for this reason has to be estimated from on site spot measurements. At least few manual traffic counts should be always taken to assure the quality of data obtained from automatic traffic networks.
3.4.2. Emissions
All street canyon models require vehicle emission factors (e.g. g/km per single vehicle) or emission rates (e.g. g/km per hour) as input, although some operational models (e.g. CAR International) might include default emission factors for certain countries. In certain models (e.g. AEOLIUS), separate emission factors for small and large vehicles need to be specified. The emission rates in a street can be derived from the traffic volumes and the composite emission factors of the pollutants. A number of methodologies and models may be applied to determine the appropriate fleet-average emission factors.
The CORINAIR working group (sponsored by the European Commission) developed a methodology for calculating emissions, including appropriate emission factors, from road traffic (Eggleston et al., 1993). The methodology was transformed into the computer program COPERT (Ntziachristos and Samaras, 1997; 2000). The same methodology was adopted by ADEME (1998) to develop IMPACT, a road traffic emission model which quantifies fuel consumption and atmospheric releases of a specified vehicle fleet in a given year in France. Emissions are calculated for two vehicle operating modes: hot and cold start. The required input parameters are traffic composition, average vehicle speed, length and slope of the road segment of interest. In addition, the month of the year is used to estimate average ambient temperatures, which are further used for calculating evaporative and cold running emissions. The model provides default values for the average travelling distance and the fraction of this distance run with a cold engine in France.
The U.S. EPA has developed and regularly revised MOBILE, which is also a mobile source emission factor model. MOBILE distinguishes moving vehicles into three operating modes: cold start, hot stabilised and hot start. The model inputs include the vehicle miles travelled by each specified type of vehicle, ambient temperature, terrain altitude, calendar year, average vehicle speed, etc. Default values applying to the U.S. vehicle fleet are provided within the model. A related EPA model, PART5. can be used to calculate emission factors for particulate matter. In California, the Air Resources Board's EMFAC model is used in place of MOBILE.
An important aspect that differentiates MOBILE from COPERT is that the latter bases its assumptions on fuel sale statistics, while MOBILE assesses annual mileage accumulation rates by using data from traffic surveys. Other significant differences lie in the way the two models account for the effect of low ambient temperatures and cold start emissions (Zachariadis and Samaras, 1999). A detailed comparative analysis of MOBILE5a and COPERT was presented by Samaras and Zachariadis (1994).
The German MOBILEV emission model can also be used to calculate yearly or hourly average emissions for a single street or a street network using available information on emission factors, traffic mode, street characteristics and vehicle fleet composition. Congested traffic and cold start emissions are also taken into account. Casella Stanger in asociation with AEA Technology developed a spreadsheet model for NOX and PMio emission factor on behalf of the U.K. Department of the Environment (DEFRA), to assist local authorities in the air quality review and assessment process. Finally, the protocol used by Buckland and Middleton (1999) can be applied for estimating composite emission factors for most regulated pollutants in the U.K. This methodology is based on predefined emission factors specific to each vehicle category. The fleet composition is then used to derive a composite emission factor for the road segment of interest.
Although this list of models is not exhaustive, it gives an idea of the existing emission calculation methodologies and the required inputs. It should be stressed that emission factors have to be regularly updated to reflect changes in fuel standards, vehicle fleet composition and engine technology (Stedman et al., 2001). It is generally recognised that emission factors represent one of the most important source of uncertainty in modelling traffic pollution (Kiihlwein and Friedrich, 2000).
3.4.3. Meteorological data
The amount of required meteorological information for air quality modelling is proportional to the sophistication of the selected model. Simple models for screening applications (e.g. CAR and AEOLIUS Screen) only require the average wind speed over a period of time, assuming that there is no prevailing wind direction. Relatively more sophisticate street canyon models (e.g. OSPM and AEOLIUS Full) require time series of wind speed and direction for the dispersion calculations, ambient temperature and global radiation for the photochemistry algorithm, and (in some cases) atmospheric pressure for unit conversion. In addition to this information, CFD codes (e.g. PHOENICS) require certain specifications concerning atmospheric turbulence and wind profiles. Finally, Gaussian models (e.g. CALINE4) may need additional information on atmospheric stability and mixing height, although these parameters generally do not play an important role in pollutant dispersion within urban canyons. This is because the mechanical turbulence produced by building and vehicles outbalances the influence of the atmospheric turbulence (Leisen and Sobottka, 1980).
It should be remembered that meteorological data obtained simultaneously at different weather stations located within few kilometre distances from each other might differ significantly, especially for short averaging periods. When the first street canyon models were developed few decades ago, it was assumed that the local roof-top wind information needed as an input would not be generally available, and airport ten-meter wind data would have to be used. For this reason, empirical expressions relating airport and local winds were derived (Johnson et al., 1973). However, recent research studies have shown that model simulations carried out using airport winds generally produce lower and less accurate air quality predictions compared to those produced using local wind data (Manning et al., 2000).
3.4.4. Street geometry
Again, more sophisticated models require a larger amount of input information. Simple models may only need the height and the width of the canyon as input (e.g. AEOLIUS Screen), or just the type of the street and the distance between simulated receptor and road axis (e.g. CAR). Semi-empirical models (e.g. OSPM) may additionally require the length and the orientation of the canyon and allow for some gaps between the buildings. A surface roughness coefficient might be also provided by the user. Although there are both experimental and theoretical methods for estimating the roughness length of an urban surface (Pal Arya S., 1988), arbitrary values (» 0.6 m) are often used.
Relatively simple mathematical models are generally not able to capture the details of the urban canopy (e.g. trees, slanted building roofs, balconies, parked cars, etc.), which might have a significant influence in small-scale pollutant dispersion within street canyons (Gayev and Savory, 1999; Rafailidis, 2000). By contrast, CFD models are able to closely reproduce the details of the urban canopy, if the necessary input information is available. Four main types of boundary conditions imposed to the physical limits of the simulated area have to be specified: (a) the walls, (b) the inlets, (c) the outlets, and (d) the planes of symmetry (i.e normal velocities are set to zero at a symmetry boundary and the values of all other properties just outside the solution domain are equal to their values just inside the domain). Most commercial CFD codes provide the necessary graphical and numerical tools for treating complex street configurations, including fixed and moving obstacles (Theurer, 1999; Venetsanos et al., 2001 ).
3.4.5. Background concentrations
Even the simplest urban canyon models require a background pollution value as input, to account for the fraction of the pollutant that is not emitted within the simulated street. The urban background concentration can be defined in several ways. Ott and Eliassen (1973) suggested the existence of an urban CO background concentration as a relatively constant concentration that would be observed at a number of locations throughout the city, providing that the observer was at least 200 feet from the nearest street. Other authors have suggested the use of roof-top measurements as an estimate of the urban background levels. An alternative approach is to simulate the entire urban area using a larger scale model in order to determine background levels contributed by non-localised sources. The main disadvantage of this method is that it requires additional input data, which are subject to uncertainties (Cooper, 1987).
Berkowicz (2000b) developed a simple model for urban background pollution that can be used in combination with OSPM. ApSimon et al. (2001) adopted ADMS-Urban within the integrated assessment model USIAM to define contributions from different sources at different background receptor locations. CAR includes a simple algorithm for deriving urban background using the regional background (i.e. background due to distant sources) and the diameter of the built-up city area. Nevertheless, the commonest (and more reliable) modelling practice is to obtain background concentrations from measurements at urban locations that are not directly affected by local sources.
3.5.1. Deposition and resuspension
The mass of total suspended particles (TSP) and PM|0, usually measured in fixed monitoring stations, is dominated by the coarser fraction of airborne paniculate matter. The PM10 fraction may be transported over long distances (Vignati et al.. 1999). It is unlikely for an all-purpose street canyon model to be able to reproduce atmospheric aerosol concentrations measured on the kerbside, unless long range transport, local and regional non-traffic sources, relative humidity, deposition and resuspention processes are adequately taken into consideration.
The number of measured particles is dominated by the smaller fraction of P M2j known as ultra-fine particles (i.e. aerodynamic diameter < 100 nm). Significant correlation at street level was observed between traffic-related gases (NOX, CO) and ultra-fine particle numbers detected in a street canyon in Copenhagen (Wâhlin et al, 2001), indicating that traffic was their major source in the urban air. Ultra-fine particles are generally expected to behave like inert gases within short distances from their sources. Therefore, their concentration may be successfully calculated using urban canyon models (e.g. OSPM) originally developed for gaseous pollutants (Le Bihan et al., 2001).
3.5.2. Chemistry
Simple street canyon models (e.g. STREET-SRI) may only calculate concentrations of passive compounds (e.g. CO). On the other hand, models that are used in regulatory applications need to take fast photochemical reactions into account in order to calculate NON concentrations. OSPM uses a simplified chemistry algorithm to account for the transformation of reactive species (i.e. NOX and O3) inside a street canyon. AEOLIUS includes a subroutine that calculates statistically N O T from NON concentrations by means of an empirical curve-fit formula derived from measurements in London (Derwent and Middleton. 1996). In that case, a maximum ratio is set to prevent the NO2 exceeding 2 5 % of the NOX at high concentrations. CAR uses an empirical relationship derived from the more elaborated TNO model to calculate street-level NO2 concentrations, depending on O3 levels, the fraction of total NOX directly emitted as NO2, and the type of the street. Final!\. CALINE4 includes the Discrete Parcel Model forNOx chemistry.
General-purpose CFD models are only able to provide concentrations of inert pollutants, since they do not usually take photochemistry into account. However, specially designed microscale models may combine CFD codes with simplified chemistry algorithms. For example, WinMISKAM adds to MISKAM a simple NO-NO2 conversion model. A simple photochemistry algorithm is also implemented in the street and neighbourhood scale MICRO-CALGRID model, which is linked with MISKAM, as an alternative to the full chemistry scheme implemented in the urban scale CALGRID model (Stern and Yamartino. 2001).
3.5.3. Vertical profiles
Relatively simple street canyon models (e.g. AEOLIUS, CAR, etc.) only calculate street-level pollutant concentrations, without giving the user the possibility of choosing the height of the simulated receptors. It has been suggested by several authors (Capannelli et al., 1977; Huang, 1979; Dabberdt and Hoydysh, 1991; Zoumakis, 1995; Vardoulakis, 2002a) that the vertical concentration profiles in a street canyon generally satisfy a law of exponential reduction with height, although more complex patterns depending on the side of the street, the distance from the walls, and the small-scale features of the buildings may be also observed (Kastner-Klein and Plate, 1999; Jicha et al., 2000). The exponential redaction law might not be applicable to traffic-related substances with very short chemical lifetime (e.g. NO, NO2). It has been experimentally demonstrated (Vakeva et al., 1999) that NO2 concentrations may even increase with height within a street canyon, when the weather conditions favour photochemical activity.
4. Street canyon studies
Several modelling and experimental field studies aiming to establish pollutant dispersion and transformation patterns within street canyons have been carried out in the past. Depending on their objectives, different modelling and monitoring techniques have been adopted. Some of these studies were purely experimental, which means that they were exclusively based on full- and/or reduced-scale measurements. At the other end of the spectrum, some purely theoretical studies mainly focusing on the investigation of different wind flow and pollutant dispersion regimes using mathematical models can be also found in the literature.
Most commonly, street canyon studies combine both mathematical modelling and experimental work. They may follow two different research approaches. The first one is based on the use of relatively simple parametric models and data obtained from field and/or wind tunnel experiments. Usually, the objective of this kind of studies is to determine the spatial and temporal variability of roadside air pollution, validate operational models, estimate population exposure, etc. The second approach is based on the use of more sophisticated CFD models and experimental data from wind tunnel and/or field measurements. The objective of these studies is usually to obtain a detailed description of the wind and concentration fields within the urban canopy under well-defined dispersion conditions.
Recently, the European research network TRAPOS (Optimisation of Modelling Methods for Traffic Pollution in Streets) gave new insights in a number of street canyon related issues: (a) the influence of moving vehicles on pollutant dispersion and turbulence in urban streets (Kastner-KIein et al., 2000 and 2001; Vachon et al., 2001); (b) the thermal effects on flow and dispersion within street canyons especially under low wind conditions (Kovar-Panskus et al., 2001a; Louka et al.. 2001); (c) the sensitivity of flow and turbulence characteristics to the geometry of the street and its surroundings (Kovar-Panskus et al., 2001b; Kastner-KIein and Rotach, 2001; Leitl et al., 2001; Chauvet et al., 2001); (d) the dispersion and transformation of traffic-related particles (Le Bihan et al., 2001; Wâhlin et al., 2001). TRAPOS included field and wind tunnel measurements, as well as mathematical simulations carried out with advanced numerical (MISKAM. CHENSI, MIMO, CFX-TASCflow) and a simpler parametric model (OSPM). A significant part of the work within the network was devoted to the inter-comparison and evaluation of these models (Sahm et al., 2001; Ketzel et al., 2001) (Fig. 3). In the following sections, representative studies covering all aspects of street canyon research are briefly discussed.
4.1. Full-scale experiments
DePaul and Sheih ( 1985; 1986) carried out a tracer gas (SF6) experiment in an urban street canyon in Chicago (U.S.A.) in order to obtain measurements of pollutant retention times and resulting concentrations within the canyon. The mean wind velocities were determined by analysing trajectories of air balloons that were released in the street. Nakamura and Oke (1988) studied the climate of urban canyons using field observations of wind and temperature from a street canyon in Kyoto (Japan). These observations were used to derive simple algorithms relating the above roof-level to the within-canyon meteorological conditions. Pfeffer et al. (1995) presented measurements of NO2, CO, benzene, soot and other atmospheric pollutants carried out in two busy street canyons in Dusseldorf and Essen (Germany), as a part of a pilot study preparing the implementation of new regulations included in the German Federal Clean Air Act. The correlation between different pollutants and the influence of the wind conditions on measured concentrations were investigated. Namdeo et al. (1999) presented results from a monitoring study on traffic-related particulate pollution in urban areas. Field measurements of airborne fine and coarse particulate matter were taken in an urban street canyon in Nottingham (U.K.) and the correlation of the observed concentrations with traffic was studied. Venegas and Mazzeo (2000) reported CO concentrations measured in a deep street canyon in Buenos Aires (Argentina).
Vertical concentration gradients of CO were observed by Zoumakis (1995) in a busy street canyon in Athens (Greece). The monitoring results were used to derive an empirical expression relating pollutant concentration and height above the ground. Gaseous pollutants (CO, NOX, O3) and aerosol particle concentrations were measured at two different heights within an urban street canyon in Lahti (Finland) by Vâkeva et al. (I999). The main objective of this study was to investigate the factors leading to the formation of vertical concentration profiles within the canyon. TSP, PMio and PM1.5 concentrations were measured in two open streets and two canyon sites in Hong Kong by Chan and Kwok (2000). These measurements showed that the dispersion of particulate matter was affected by the prevailing wind direction and the aspect ratio of the street. An exponential reduction of TSP and PM,0 with height was observed.
Vachon et al. (2000) reported results (i.e. concentration, temperature and wind fields) from a full-scale experiment carried out in a street canyon (Rue de Strasbourg) in Nantes, France. This was the first campaign of the URBCAP project, which has the aim of assessing pollutant transformation processes within the urban canopy and validating small-scale dispersion models. Finally, within the framework of the LIFE RESOLUTION project (Wright, 2001), benzene and NOi measurements were taken in four European cities (Dublin, Madrid, Paris and Rome) in order to assess pollution levels with reference to established air quality standards, optimise the design of monitoring networks, and provide experimental data to support the validation of urban dispersion models. The sampling, carried out with diffusive tubes, covered a wide range of urban and suburban locations, including a number of street canyons.
4.2. Reduced-scale experiments
Leisen and Sobottka (1980) made a comparison between field observations from two street canyons in Cologne (Germany) and wind tunnel measurements in order to investigate pollutant dispersion within urban streets and develop simulation models. Meroney et al. (1996) presented a wind tunnel study of car exhaust dispersion from street canyons in an urban environment. The main objective of this study was to investigate how pollution dispersion is affected by street geometry and particular emphasis was put on the design of a line source to realistically represent traffic emissions. The experiments were performed in the atmospheric boundary layer wind tunnel (BLASIUS) of the Meteorological Institute of Hamburg University (Germany). In a later study, Pavageau and Schatzmann (1999) investigated the concentration fluctuations in a reduced-scale urban canyon simulated within BLASIUS. Experimental data sets of wind tunnel measurements carried out in this facility for the validation of microscale dispersion models are available on the Internet (http://www.mi.uni-hamburg.de).
The differences between reduced- and full-scale experiments were illustrated by Liedtke et al. (1999), who compared field measurements obtained in a street canyon in Hanover (Germany) with wind tunnel results. The effect of using a simplified model of a street canyon in the wind tunnel was studied by taking measurements using different scaled models that included various levels of detail of the real canyon geometry. Significant differences were found in the results. In a later study, Schatzmann et al. (2000) showed how wind tunnel data can be used to supplement and enhance the value of field measurements for model validation purposes.
Rafailidis (1999) investigated the influence of atmospheric thermal stratification on urban street canyon ventilation in the EnFlo wind tunnel of the University of Surrey (U.K.). The measurements indicated that stable stratification conditions result in trapping the pollutants within the canyon. In a number of other wind tunnel studies (Rafailidis, 1997; 2000), the influence of building area density and roof shape on the wind field above and inside the urban canopy were highlighted. Uehara et al. (2000) also investigated the effects of thermal stratification on the wind flow in and above urban street canyons using the atmospheric diffusion wind tunnel at the Japanese National Institute for Environmental Studies (Ogawa et al.. 1981 ). The results showed that the wind vortex within the canyon becomes weaker when the atmosphere is stable. Kastner-Klein and Plate (1999) presented results from tracer dispersion experiments performed in a neutrally stratified atmospheric boundary layer wind tunnel in the Institute of Hydrology and Water Resources of the University of Karlsruhe (Germany). The influence of systematic parameter variation (i.e. building configuration, roof shape, wind direction) on the concentration field within a street canyon was studied.
Gerdes and Olivari (1999) studied the wind and concentration fields generated within even and asymmetric street canyons under perpendicular winds using the L-2B wind tunnel of the von Karman Institute (Belgium). A strong influence of the surrounding landscape on pollutant dispersion was observed. The ratio of the height of the walls flanking the street was found to have a significant effect on the concentration patterns, while the width of the canyon was proved to be of less importance.
4.2. Parametric modelling using fieldand/or wind tunnel measurements
Johnson et al. (1973) developed STREET-SRI using data from the San Jose Street Canyon Experiment in California. Sobottka and Leisen (1980a,b) created a modified version, called MAPS, which is quite similar in form and performance with the original model. Nicholson (1975) developed a simple box model that yields street-level average CO concentrations in urban canyons under perpendicular and parallel wind conditions. Model results were proved to be in reasonable agreement with field data obtained in Frankfurt (Germany), Madison and Chicago (U.S.A.). Yamartino and Wiegand (1986) developed the CPBM using data from an extensive field monitoring programme in Bonner Strasse (Cologne). Part of the experimental data (CO and NOX) was used to evaluate the model. In the same study, the performance of CPBM (only for CO predictions) was compared with the performance of STREET-SRI and its modified version MAPS.
Hoydysh and Dabberdt (1988; 1994), and Dabberdt and Hoydysh (1991) carried out flow visualisation and tracer concentration measurements in the atmospheric boundary layer wind tunnel (ABLWT) of the Environmental Science and Services Corporation (U.S.A.). Pollutant dispersion was simulated using reduced-scale models for street canyons (both even and asymmetric) and intersections. Wind tunnel results were compared with concentrations calculated using STREET-SRI and the analytical model developed by Hotchkiss and Harlowe (1973). Finally, a simple exponential law describing vertical concentration profiles was established.
Hertel and Berkowicz (1989a) developed OSPM using measurements obtained in Jagtvej street in Copenhagen (Denmark). An intensive monitoring site was established in connection with a permanent air quality station operating in this street. A selection of the obtained wind and turbulence data was analysed by Nielsen (2000). OSPM has been applied 14
to several street canyons in Copenhagen, Utrecht, Oslo, Helsinki, Beijing and other major cities (Berkowicz et al., 1996; Hertel and Berkowicz, 1989b; 1989c; Kukkonen et al., 2001; Fu et al., 2000). Based on the same principles, Buckland ( 1998) formulated AEOLIUS, which has been mainly used in the UK (Manning et al., 2000). Other authors (Sacré et al.,
1995; Vardoulakis et al., 2002b) have presented slightly modified versions of OSPM.
Kono and ko (1990a) presented the OMG VOLUME-SOURCE model, a microscale dispersion model that estimates concentrations of traffic-related pollutants in an urban area within 200 m from the roadside. The model parameters were determined using experimental data from five locations in Osaka (SF6 was released as a tracer gas). Model results were compared by the same authors (Kono and Ito, 1990b) with concentrations calculated using three line source dispersion models, namely the JEA model (Japan Environmental Agency, 1982), the TOKYO model (Tokyo Metropolitan Government. .1983). and the HIWAY-2 model (Peterson, 1980).
Qin and Kot (1993) took measurements of CO and NOX at different heights and distances from the kerb within three asymmetric street canyons in Guangzhou City (China). STREET-SRI and a Gaussian plume model were used in this study to obtain CO and NOX estimates, which were found to be in reasonable agreement with the observed values. Lanzani and Tamponi (1995) presented the microscale Lagrangian particle model GEM for the atmospheric dispersion of primary pollutants in the urban canopy. In the same study, GEM was validated against field measurements and compared with CPBM and STREET-SRI. The STREET-SRI mode! was also used in two independent studies in Argentina to calculate CO concentrations in street canyons in Cordoba (Stein and Toselli, 1996) and Buenos Aires (Bogo et al., 2001). A reasonably good agreement was found between measurements and predictions.
Hargreaves and Baker (1997) developed a Gaussian puff model, called PUFFER, to simulate the dispersion of vehicular pollutants in urban street canyons. This model, which explicitly takes into account vehicle induced turbulence, enables the user to investigate realistic transient situations such as traffic congestion and non-steady above canyon wind fields. A short sensitivity analysis and a comparison with STREET-SRI were also included in the same study. Gualtieri and Tartaglia ( 1998) developed a comprehensive air quality model, including traffic, emission and dispersion sub-models, for assessing pollutant concentrations in urban areas. A semi-empirical street canyon algorithm based on STREET-SRI and field measurements from Firenze (Tartaglia et al., 1995; Gualtieri and Tartaglia, 1997) was included in this model, which was finally integrated in a Geographic Information System (GIS).
Hassan and Crowther (1998a) developed a single box model for calculating first estimates of pollutant concentrations within urban canyons. The model parameters were derived using field CO measurements taken in Hope Street, Glasgow (U.K.)- Furthermore, Hassan and Crowther (1998b) used PHOENICS to simulate wind flow and pollutant dispersion within the same canyon. The accuracy of the two-dimensional steady state numerical simulations was examined by comparing the predicted results with the field measurements.
Micailef and Colls (1999) developed a semi-empirical emission-dispersion model for predicting the temporal and spatial distribution of airborne particulate matter in street canyons. This model called SLAQ includes emission modules, meteorological pre-processors, modules for within-canyon processes, dispersion modules, and modules accounting for external influences. The dispersion module is mainly based on CPBM. Model features include a correction for the heat island effect, dry and wet deposition, particle settling, etc. SLAQ was evaluated against measurements obtained in a street in Loughborough (U.K.) using the automated near real time Kinetic Sequential Sampling (KSS) system (Colls and Micailef. 1999).
Coppalle et al. (1999, 2001) measured NOX and CO concentrations at different background and kerbside locations, including a street canyon (Rue Crevier), in a medium size French city (Rouen) during four weeks in winter. Using the obtained experimental data for parameterisation, Coppalle (1999) developed a simple mathematical model that calculates the vertical pollutant distribution inside a street canyon under low wind conditions. Vardoulakis et al. (2000, 2002a) carried out field experiments in two even (Bd. Voltaire and Rue de Rennes) and one asymmetric (Av. Leclerc) street canyons in Paris, France. Three semi-empirical dispersion models (STREEET-SRI, OSPM and AEOLIUS) were tested using the observed CO and benzene values (Fig. 4). Furthermore, an empirical relationship for calculating vertical concentration profiles within the streets was derived.
In the framework of AUTO-OIL II programme (Skouloudis, 2000), a large number of air quality modelling simulations were carried out in order to assess the compliance with the new EU air quality standards for NO2, O3, CO, benzene, and PM|0. An advanced methodology was established, incorporating models of different spatial scales, to provide air quality simulations down to street level. Two urban canyon models, namely OSPM and MICRO-CALGRID, were evaluated using measurements from Viale Murrilo (Milan) and Schildhorn Strasse (Berlin).
